{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:OFXK4YJZV636PJQVHLWDIXGNRP","short_pith_number":"pith:OFXK4YJZ","schema_version":"1.0","canonical_sha256":"716eae6139afb7e7a6153aec345ccd8be021244bafcde8a8c024486d496242bb","source":{"kind":"arxiv","id":"2607.07508","version":1},"attestation_state":"computed","paper":{"title":"Single-Rollout Asynchronous Optimization for Agentic Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Jie Tang, Yujiang Li, Yuxiao Dong, Zhenyu Hou","submitted_at":"2026-07-08T15:02:19Z","abstract_excerpt":"Reinforcement learning (RL) is becoming increasingly important for post-training large language models (LLMs). Previous RL pipelines for LLMs were mostly synchronous and batch-interleaved, which is inefficient for long-horizon agentic tasks. Recently, asynchronous RL has emerged as a more efficient alternative by updating the model as rollouts arrive. However, existing asynchronous RL systems often emphasize throughput, while leaving training stability and task effectiveness largely underexplored. For example, a key challenge is that group-wise sampling in the widely-used GRPO framework does n"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2607.07508","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-08T15:02:19Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"8f2a275f3755ef0674d8e6d9770677b0961bb54c288140853ab58e8d5f3da9cb","abstract_canon_sha256":"d2eff236798f1e53ae28065c651f03618bda8753cb7286a98ed109cfaffc295e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-09T01:20:27.611757Z","signature_b64":"wONyD9SSlEZoqjSrRcqGH0H3mx44vhbj7jyZggpMgJOv1yi58+kpdeWRawW3PXoCzqg4z5bb7OuZQ3vIx/FjBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"716eae6139afb7e7a6153aec345ccd8be021244bafcde8a8c024486d496242bb","last_reissued_at":"2026-07-09T01:20:27.611346Z","signature_status":"signed_v1","first_computed_at":"2026-07-09T01:20:27.611346Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Single-Rollout Asynchronous Optimization for Agentic Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Jie Tang, Yujiang Li, Yuxiao Dong, Zhenyu Hou","submitted_at":"2026-07-08T15:02:19Z","abstract_excerpt":"Reinforcement learning (RL) is becoming increasingly important for post-training large language models (LLMs). Previous RL pipelines for LLMs were mostly synchronous and batch-interleaved, which is inefficient for long-horizon agentic tasks. Recently, asynchronous RL has emerged as a more efficient alternative by updating the model as rollouts arrive. However, existing asynchronous RL systems often emphasize throughput, while leaving training stability and task effectiveness largely underexplored. For example, a key challenge is that group-wise sampling in the widely-used GRPO framework does n"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.07508","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2607.07508/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2607.07508","created_at":"2026-07-09T01:20:27.611406+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.07508v1","created_at":"2026-07-09T01:20:27.611406+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.07508","created_at":"2026-07-09T01:20:27.611406+00:00"},{"alias_kind":"pith_short_12","alias_value":"OFXK4YJZV636","created_at":"2026-07-09T01:20:27.611406+00:00"},{"alias_kind":"pith_short_16","alias_value":"OFXK4YJZV636PJQV","created_at":"2026-07-09T01:20:27.611406+00:00"},{"alias_kind":"pith_short_8","alias_value":"OFXK4YJZ","created_at":"2026-07-09T01:20:27.611406+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.01418","citing_title":"Reusing Rollouts under Policy Lag: Prefix-Normalized Policy Optimization for LLM Reinforcement Learning","ref_index":4,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OFXK4YJZV636PJQVHLWDIXGNRP","json":"https://pith.science/pith/OFXK4YJZV636PJQVHLWDIXGNRP.json","graph_json":"https://pith.science/api/pith-number/OFXK4YJZV636PJQVHLWDIXGNRP/graph.json","events_json":"https://pith.science/api/pith-number/OFXK4YJZV636PJQVHLWDIXGNRP/events.json","paper":"https://pith.science/paper/OFXK4YJZ"},"agent_actions":{"view_html":"https://pith.science/pith/OFXK4YJZV636PJQVHLWDIXGNRP","download_json":"https://pith.science/pith/OFXK4YJZV636PJQVHLWDIXGNRP.json","view_paper":"https://pith.science/paper/OFXK4YJZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.07508&json=true","fetch_graph":"https://pith.science/api/pith-number/OFXK4YJZV636PJQVHLWDIXGNRP/graph.json","fetch_events":"https://pith.science/api/pith-number/OFXK4YJZV636PJQVHLWDIXGNRP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OFXK4YJZV636PJQVHLWDIXGNRP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OFXK4YJZV636PJQVHLWDIXGNRP/action/storage_attestation","attest_author":"https://pith.science/pith/OFXK4YJZV636PJQVHLWDIXGNRP/action/author_attestation","sign_citation":"https://pith.science/pith/OFXK4YJZV636PJQVHLWDIXGNRP/action/citation_signature","submit_replication":"https://pith.science/pith/OFXK4YJZV636PJQVHLWDIXGNRP/action/replication_record"}},"created_at":"2026-07-09T01:20:27.611406+00:00","updated_at":"2026-07-09T01:20:27.611406+00:00"}